Over 11 million red blood cell units are transfused in the United States annually. Despite the common use of red blood cell transfusions, the threshold for transfusion has not been adequately evaluated. A decade ago the standard of care was to administer a peri-operative transfusion whenever the hemoglobin level fell below 10 g/dl (the "10/30 rule"). Concerns about the safety of blood, especially with respect to HIV and hepatitis, and the absence of data to support a 10 g/dl threshold led to the current standard of care to administer blood transfusions based on the presence of symptoms and not a specific hemoglobin/hematocrit level. However, there are no randomized clinical trials in surgical patients that have tested the efficacy and safety of withholding blood until the patient develops symptoms or of the "10/30" approachto transfusion and limited evidence for patients with underlyingcardiovascular disease who are at greatest risk of adverse effects from reduced hemoglobin levels. We propose to conduct a multi-center randomized trial to test if a more aggressive transfusion strategy that maintains postoperative hemoglobin levels above 10 g/dl improves patient outcome as compared to a more conservative strategy that withholds blood transfusion until the patient develops symptoms of anemia. Patients eligible for the trial will have undergone surgical repair for a hip fracture and have a postoperative hemoglobin level below 10 g/dl within three days of surgery. Only patients with cardiovascular diseasewill be entered into the study. Symptomatic transfusion strategy patients will receivered blood cell transfusions for symptoms of anemia, although transfusion is also permitted, but not required, if the hemoglobin level falls below 8 g/dl. Outcomes will include functional recovery {primary outcome: ability to walk ten feet across a room without human assistance at 60-days post-randomization), long-term survival, nursing home placement, and postoperative complications (death in hospital or within 30 days, pneumonia, myocardial infarction, thromboembplism, stroke). We will-enroll 2,600 patients over a 3.5-year period to detect a reduction in the loss of ability to walk independently from 43% to 36% (16% relative risk reduction) with power about 0.90. Also, this study will measure the frequency and 95% confidence intervals of medical errors that are important in this patient population. The medical errorsthat will be measuredare: transfusion errors (blood transfusion to the wrong patient, mislabeling of samples for type and cross match, use of whole blood instead of packed red cells), failure to use thromboembolism prophylaxis, incorrect antibiotic prophylaxis, wrong site surgery, and femoral shaft fracture.